The Research And Implementation Of Fault Diagnosis Algorithm For High-speed Train Running-gear

The running-gear of high-speed train is an important part to ensure the normal operation of the train.If the high-speed train running-gear goes wrong,it will directly affect the safe operation of trains.Therefore,to ensure the normal work of the running-gear is the premise to ensure the safety of the train operation.In order to solve the problem of low diagnostic accuracy of traditional fault diagnosis algorithms,to improve the accuracy of diagnostic algorithm is one of the important research contents in the fault diagnosis system of high-speed train for running condition diagnosis of high-speed train.The thesis proposes a fault diagnosis algorithm for the high-speed train running-gear based on multi-sensors and multi-features.The algorithm uses the train signals collected from multiple acceleration sensors installed in the running-gear as data sets.The main research work in this thesis is as follows:1.Characteristic processing of sensor signal.Propose a multi-sensor and multi-feature oriented MSMFP feature processing method:(1)feature extraction of sensor signals by using information entropy and ensemble empirical mode decomposition to establish initial feature space;(2)feature fusion SEPMFI algorithm is proposed to realize multi-feature fusion of single sensor;(3)feature combination of same feature in multi sensors is realized by MSFFC method to establish the new feature space.2.Constructing fault feature space.Use K-means clustering algorithm to calculate the fault distinguishability of features.The experimental result shows that,the fault distinguishability of new feature space is higher than that of the initial feature space.Identify and select the feature with the highest fault distinguishability to construct the fault feature space.3.Confirming fault identification algorithm.Use K-nearest neighbor,support vector machine and PSO-SVM algorithms to realize fault recognition.The experimental result shows that,the PSO-SVM algorithm has the highest fault recognition rate among the three algorithms in the fault feature space.